Light

thuml / mdd Goto Github PK

View Code? Open in Web Editor NEW

127.0 4.0 27.0 83 KB

Code released for ICML 2019 paper "Bridging Theory and Algorithm for Domain Adaptation".

Python 98.06% Shell 1.94%

mdd's Introduction

Margin Disparity Discrepancy

Prerequisites:

Python3
PyTorch ==0.3.1 (with suitable CUDA and CuDNN version)
torchvision == 0.2.0
Numpy
argparse
PIL
tqdm

Dataset:

You need to modify the path of the image in every ".txt" in "./data".

Training:

You can run "./scripts/train.sh" to train and evaluate on the task. Before that, you need to change the project root, dataset (Office-Home or Office-31), data address and CUDA_VISIBLE_DEVICES in the script.

Citation:

If you use this code for your research, please consider citing:

@inproceedings{MDD_ICML_19,
  title={Bridging Theory and Algorithm for Domain Adaptation},
  author={Zhang, Yuchen and Liu, Tianle and Long, Mingsheng and Jordan, Michael},
  booktitle={International Conference on Machine Learning},
  pages={7404--7413},
  year={2019}
}

Contact

If you have any problem about our code, feel free to contact [email protected].

mdd's People

Contributors

Stargazers

Watchers

mdd's Issues

Total loss changes to Nan somehow

Hi, thanks for sharing your nice code. I found that the total loss changes to Nan somehow and the accuracy on the target domain would drop to zero at the same time.

visda2017 accuracy

I try to experiment with visda2017 data

The paper shows 74% of the results, but my experiment is 66%

Do I have to change hyper parameters on your code to reproduce the 74% results?

Problem about accuracy

Hello, Thank you for your reply about accuracy.
Yes, it is a problem about pytorch version.
In pytorch1.0, I just change the code:
accuracy = torch.sum(torch.squeeze(predict).float() == all_labels) / float(all_labels.size()[0]) in train.py to --->
accuracy = Tensor.cpu(torch.sum(torch.squeeze(predict).float() == all_labels)).numpy() / float(all_labels.size()[0])

It will run rightly.
Thank you for your reply again.

Legacy autograd function

I got an error on:
RuntimeError: Legacy autograd function with non-static forward method is deprecated. Please use new-style autograd function with static forward method. (Example: https://pytorch.org/docs/stable/autograd.html#torch.autograd.Function)

Could you spent some time to fix it?

Problem about accuracy

Hello,thank you for your sharing.
I have a problem. When I run your codes, the eval accuracy is always 0. Just as bleow.
Cloud please tell me something about it?

init_lr: 0.004
lr_scheduler: {'gamma': 0.001, 'decay_rate': 0.75, 'type': 'inv'}
optim: {'type': 'sgd', 'params': {'momentum': 0.9, 'nesterov': True, 'weight_decay': 0.0005, 'lr': 0.004}}
start train...
Train iter: 0%|▍ | 500/100000 [03:21<11:58:55, 2.31it/s]
{'accuracy': tensor(0, device='cuda:0')} | 4/16 [00:01<00:05, 2.14it/s]
Train iter: 1%|▋ | 1000/100000 [06:48<10:18:59, 2.67it/s]
{'accuracy': tensor(0, device='cuda:0')} | 8/16 [00:03<00:03, 2.60it/s]
Train iter: 2%|█▏ | 1500/100000 [10:14<9:52:52, 2.77it/s]
{'accuracy': tensor(0, device='cuda:0')}▊ | 12/16 [00:04<00:01, 2.74it/s]
Train iter: 2%|█▌ | 2000/100000 [13:31<8:34:56, 3.17it/s]
{'accuracy': tensor(0, device='cuda:0')} | 0/16 [00:00<?, ?it/s]
Train iter: 2%|█▊ | 2500/100000 [16:58<12:21:09, 2.19it/s]
{'accuracy': tensor(0, device='cuda:0')} | 4/16 [00:01<00:06, 1.86it/s]
Train iter: 3%|██▎ | 3000/100000 [20:25<9:56:52, 2.71it/s]
{'accuracy': tensor(0, device='cuda:0')} | 8/16 [00:03<00:03, 2.64it/s]
Train iter: 4%|██▋ | 3500/100000 [23:53<9:39:06, 2.78it/s]

Where is the 'All proofs can be found in Appendices A-C.'?

Just can not find it.

Please post issues at https://github.com/thuml/Transfer-Learning-Library

This repo is obsolete. Please describe your issues at https://github.com/thuml/Transfer-Learning-Library

Recommend Projects

React

A declarative, efficient, and flexible JavaScript library for building user interfaces.
Vue.js

🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript

TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
TensorFlow

An Open Source Machine Learning Framework for Everyone
Django

The Web framework for perfectionists with deadlines.
Laravel

A PHP framework for web artisans
D3

Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

javascript

JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
web

Some thing interesting about web. New door for the world.
server

A server is a program made to process requests and deliver data to clients.
Machine learning

Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Visualization

Some thing interesting about visualization, use data art
Game

Some thing interesting about game, make everyone happy.

Recommend Org

Facebook

We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft

Open source projects and samples from Microsoft.
Google

Google ❤️ Open Source for everyone.
Alibaba

Alibaba Open Source for everyone
D3

Data-Driven Documents codes.
Tencent

China tencent open source team.